THE HISTORY AND SHAPE OF THE TEST-PIECE. 325 lower elastic ratio, a better elongation, and a lower reduction of area. Some of these characteristics may be due to differences in finishing temperature, but the data on elastic limits show that the pieces were all rolled at nearly the same degree of heat, and such small variations are not sufficient to account for the increase in the elongation. This variation in physical qualities, as produced by differences in diameter, has been discussed by Appleby.* In common with others, he makes the fundamental mistake of rolling all the bars to one size, viz., 1-J inches in diameter, and turning the test specimens from these bars. A test-piece of one-half inch in diameter thus obtained will be merely the core or center of the original bar, and will be inferior both chemically and physically. On the one hand it embraces the area of maximum segregation, while on the other it has not undergone the compression that the exterior of the bar has received in the rolls, and. a comparison of the bars is invalid. The method, which I have employed, of comparing rolled bars of different sizes in the form in which they left the rolls, also presents complicating conditions, inasmuch as the effect of work is not the same on large and on small sections, but it has the advantage that it represents actual conditions. Siio. XVIi.—Influence of fhe width of the test-piece.—Conclusive testimony that variations in the elongation may be due solely to the cross-section of the test-piece is furnished by Table XVI-L, which gives the results obtained in breaking strips of different width when the pieces were cut side by side from the same plate. No comparison can be made between the different thicknesses, since the individual heats were not the same, but in the matter of • widths the case is otherwise, for every heat in the group was tested in all the widths, the bars from each heat being cut from the-same small strip of plate, and this should give a valid basis of comparison. The conclusions are as follows: (1) Variations in the width of the test-piece have little effect upon the ultimate strength per square inch. (2) They probably have little influence upon the elastic limit. The narrowest pieces show a decided increase, but this needs cor-roboration. The three-inch" pieces were pulled at the works of the * Proc. Inst. Civil Eng. (England), Vol. CXVIII, pp. 395-417.